Weinroth: There is a strong belief among social software enthusiasts and developers that tagging and folksonomies will ultimately replace folders and other hierarchy-based categorization systems on a wholesale basis. Part of the value of folders and hierarchies is that they provide an exact "location" for articles and other content. What are your thoughts on the future roles of tagging and hierarchies relative to each other and in the context of day-to-day computing for average technology users?

Shirky: From my point of view, the basic premise here is flawed. There is only value in exact location when using systems, whether mental or electronic, that require that kind of location to function, so its a chicken/egg problem. Preserving the "value" of exact location really means maintaining the anachronistic assumptions embedded in systems that presume hierarchal organization in the first place.

Gutenberg's invention of movable type gave us a world where a book could be created in less time than it took to read it. From the resulting explosion of books came the need to organize them systematically, and the categorization systems that were developed in response -- Dewey Decimal and the Library of Congress are the two most familiar -- were built around exact location because books _have_ an exact location, as all physical objects do.

Now we have data that does not have an exact location, or, rather, can have multiple exact locations, and it would be stupid to persist in applying old tools to the current world. Arguing that we should give up on non-hierarchal organization because all the systems we developed for the physical world use it is like the RIAA arguing that because music used to be expensive and inefficient to distribute, it should always remain so.

As for the future, hierarchies have a bunch of really good mathematical properties that make them useful, like being able to hold whole collections of things under one node and handle them as a unit, but those properties don't map well to the user view of the content, which is much fuzzier and multi-variate. We can get the best of both worlds -- crunchy hierarchal goodness wrapped in a creamy tag-based access interface -- without ever forcing the users to think hierarchically, which will be boatloads better than spending countless cycles training users to think/like/this.

Lebkowsky: Some people will always want to use the folders and files metaphor, and it'll probably persist as a way to organize data on a storage device like a local hard drive. On the web I think we've already abandoned hierarchies of folders in favor of links to various resources. Links are flexible, you can organize 'em many ways. Tags let us aggregate links categorically, and the categories are not imposed or enforced by some external authority. We make our own, and in systems that accommodate social tagging, we share both categories and the resources they link to. This is a great way to categorize data we find on the web, but it's a bit chaotic, and where links and resources are shared the volume of data can feel unmanageable. I think we'll see more and more ways to use tags, and more (and more manageable) ways to sort and filter those tags and the resources they reference. I think of tags as another good way to surf the web, though, rather than a way to find specific content. We'll use our increasingly sophisticated search engines for that.

Weinroth: So far tagging is emerging in several high-level application areas such as vertical search, social bookmarking and personalized search. Assuming tagging emerges and establishes itself as key component of the web experience for people other than early adopters and enthusiasts, what do you think will be tagging's most valued and relied upon application areas?

Shirky: Users only ever ask themselves "Where is that information?" as a proxy to their real desire, which is to be able to use the information somehow -- the "Find" step is simply a necessary evil. The most valuable aspect of tagging is to make the Find step more transparent.

And this is really two changes in one. The first, and obvious one, is 'Search by tag.' This is the most obvious because it fits best with our current Google-flavored model for finding things on the Web, and it will create some incremental value for some searches, perhaps especially of blogs.

The second is keeping found things found. We have come from a land of information scarcity, and the assumptions from that world break in this one. In a world of information scarcity, the assumption is that after a user has some piece of information, it's problem solved. Once the library has given the patron the book, they should have no trouble keeping track of it.

In a world of information abundance, though, it's merely problem transferred, because now the user has so much information that they have become an accidental archivist, and everyone has the same two, bad
strategies: get rigid about a categorization scheme for local files, and use some combination of bookmarks and common search strategies. (How often have you had to re-find something by recreating a particular Google search?)

Tags put post-industrial strength tools in the hands of ordinary users, allowing them to manage the increasingly large corpus of things they've already found once, allowing them to re-find them easily.

Lebkowsky: I'm interested in the prospects for using tags as filters, the various ways that can work. On a social networking site, you might have an ability to see tags for specific levels of your network, with the assumption that those tags would relate to content of interest from trusted sources. I've wanted to see someone build a site where you could make quick notes and tag them, a way to categorize virtual "scraps." And I think that every publishing site will eventually adopt some system to allow subscribers to tag any item - articles, pictures, blog posts, etc.

Weinroth: In an open model, web users will have highly divergent tagging behavior; some will use tags to categorize content (like folders), and others will use tags in a purely annotative or descriptive sense (like keywords or notes). Users will also spell and phrase tags differently. Do you believe there are any differences in user value between a tag cloud that is very consistent (people tag the content in a highly uniform manner) and one that is highly fractionalized (people tag the content in many different ways)?

Shirky: First of all, almost no one tags like folders. Why bother? If you wanted a system like folders, you'd use folders. The 'tag X must contain all and only examples of tag Y' constraint is utterly arbitrary, as it reflects neither things that are true about the world (as with a web page that is about both design philosophy and programming), nor is it a requirement of the tagging system. Using tags like folders is like drawing Venn diagrams with no overlapping circles -- possible, but so destructive of the value of the system as to make the effort pointless.

I know I keep harping on this, but it's such a common error, and its the biggest single block to understanding tagging. Whenever you find someone tagging like folders, you are seeing a failure of the imagination at work.
In an electronic world, there is almost no value in preserving expectations, even metaphorical ones like folders, held over from a world where things have to be in places. In a thingless, placeless world, that idea is simply irrelevant, as are the management systems that grew up around it, and users are reacting to this, tagging away without regard to previous, and inoperable, constraints.

As for uniform tagging, that can only work in situations where there is enough force to expend making the users behave uniformly, and where it is worth expending that force. For example, the Diagnostic and Statistical Manual, (DSM-IV) used by American psychiatrists and psychologists, provides a relatively standard way to diagnose mental disorders. It only works as well as it does, however, and that not perfectly, because it is produced by the American Psychiatric Association, which body can exert considerable force over its members, and DSM-IV is only to be used by the members.

The amount of human cost, in other words, in creating an enforcing uniformity is so high that attempts at such uniformity will fail in most cases. Fortunately, tagging allows for degenerate cases such as alternate spellings and phrasing. This would be a problem if there were only one tagger, responsible for a large group of users, but with every user a tagger, the loss stemming from degenerate cases quickly shrinks, while the value from multiple points of view grows.

The more interesting use of uniformity is inventing a tag as a shibboleth, a word or phrase that allows a group to coordinate it's tagging efforts.

There is a group looking at a particular subset of technology for the non-profit world on del.icio.us, and they tag everything relevant to the group effort with 'nptech.' This is a tag unlikely to arise at random, so the effect is to make nptech a group handle, the one tag that will return all links tagged by the group, _without requiring uniformity of classification in all dimensions._ Group membership, in other words, rather than homogenized worldview, becomes the side of just enough uniformity to create both coordinated group effects and still preserve the value of multiple points of view.

Group tagging is the next big frontier, or one of them. This work will include, of course, spam defenses like "Only show me things tagged 'design' by people I know, or people they know." As we know from years of bitter experience, it's impossible to create a source of group-created value with low barriers to entry without also inviting system gaming.

Lebkowsky: In social tagging environments, I think there’ll be a tendency to conform to the most common categories, so you’ll have more consistency, and that makes sense for an environment where the goal is to share whatever resources are being tagged – you look for a common vocabulary and a shared set of meanings, so there’s value in consistency. However one value of tagging is that you can come up with a categories that are more nuanced, so I think you’ll still see what might appear to some to be redundant categories that have value because they acknowledge subtle differences. Personal tagging may be less uniform because your tags have to mean something only to you, based on the way you think. If you’re very consistent and orderly you’ll tag that way in a personal context, but if you’re more – what’s a good word, maybe *fractal* - in your thinking, you’ll may be less consistent because that works for you.

Weinroth: As tagging gains interest among developers the issue of standards and protocols, in a purely technical sense, is beginning to be discussed. What do you see as key opportunity areas for applying code level technology standards in tagging? Are there any important disadvantages?

Shirky: The context of a tag is critical -- users tag differently on del.icio.us than on Flickr, so treating tags as purely atomic elements strips them of much of their value. A 'tag molecule', so to speak, includes not just the tag and the URI it pointed to but the users, timestamp, service it was derived from, and other tags by that user pointing to the same URI. Getting that bit of encoding right seems to me to be the essential issue.

Lebkowsky: I see one opportunity, but it’s big: I think it’d be cool if tags could be created anywhere and be meaningful everywhere. You might do this with a standard format like xFolk. Technorati came up with a simple solution, a rel (or relationship) attribute of “tag” that you can use in a link reference, e.g. Blog links the term Blog to Technorati’s index of items with the Blog tag, and the rel attribute tells Technorati to regard the link as a tag, and that the blog post should be added to Technorati’s index for “Blog.” The question is whether any of these formats will be widely adopted, and if you adopt one standard, is it flexible enough to support further evolution of tagnology. (Sorry about that pun, I couldn’t resist.)

Weinroth: In your opinion, what are some of the most interesting tagging-related projects being worked on that readers may not have heard of yet?

Shirky: Maps.

There are so many cool 'tag this location' efforts going on it's hard to keep track of them -- Platial, FoundCity, Tagzania, and so on. All these efforts take the basic intuition of tagging -- URLs are a good thing to attach labels to -- and applies it to the map world -- locations are a good thing to attach labels to. And because tagging is so flexible, the services can be used for everything from restaurant recommendations to art walks to city guides for tourists or new arrivals.

A year from now there will be an astonishing amount of value being created by tagging maps. Two years from now we'll all be shaking our heads wondering how we could have lived without that function. Three years from now, we won't remember a time when you couldn't tag maps.

Lebkowsky: I think Dinnerbuzz is a cool idea – a restaurant review multiblog with tags. Shadows lets you tag pages but there’s more – each page that’s tag has a “shadow page” where users can post comments and discussion. (The sample page they show isn’t very inspiring, but I like the idea.) Tagifieds is a tag-based bulletin board. And there’s Wists, the visual bookmarking system – which allows you to nab an image from the site as part of your bookmark. Odeo has tags for podcasts. Interesting that something as simple as tagging can trigger so much innovation, eh?